Anthropic’s head of global affairs Michael Sellitto visited Tokyo on May 15 and expressed willingness to join a cyber defense corporate coalition proposed by Japan’s ruling Liberal Democratic Party. The coalition would bring AI developers and the Japanese government together to manage risks from Mythos, Anthropic’s vulnerability-discovery AI agent that the company has kept under controlled access due to its capability to identify software flaws at scale.
The Meetings
Sellitto met with Masaaki Taira, chairman of the LDP’s Headquarters for National Cybersecurity Strategy, at party headquarters on Friday, according to Jiji Press via Nippon.com. Sellitto told Taira that Anthropic “hopes to fully cooperate with Japan’s measures against cyberattacks” and that “Japan is one of the most important countries for Anthropic.”
In a separate interview with Nikkei, Sellitto expressed a positive stance on participating in the cyber defense framework envisioned by the LDP. As the executive overseeing Anthropic’s global policy, the visit represents Anthropic’s first direct engagement with Japan’s government on Mythos cooperation.
Asked whether Japanese companies would receive Mythos access for defensive purposes, Taira told News on Japan: “I cannot comment on matters that have not yet been decided.” But he added that “there was a statement from Anthropic that the company would like to provide full cooperation wherever possible.”
Government Response
Japan’s government is moving fast. Digital Minister Masaru Matsumoto announced on May 15 that an inter-ministerial council meeting on Mythos would be held on May 18, according to the Nikkei report. The council plans to compile countermeasures for scenarios involving malicious use of the AI agent.
“We believe we have received an extremely valuable and timely set of recommendations. Based on their content, we must also formulate countermeasures,” Matsumoto stated.
The council is expected to include participants from Japan’s Digital Agency, Cabinet Secretariat, National Police Agency, Ministry of Defense, and Ministry of Economy, Trade and Industry. The agenda will likely cover legal regulation, international coordination, and operational methods for the corporate coalition.
The LDP’s cybersecurity headquarters hand-delivered a formal recommendations document to Prime Minister Sanae Takaichi on May 14, according to Nikkei. Earlier in the week, Japanese Finance Minister Satsuki Katayama met U.S. Treasury Secretary Scott Bessent and announced Japan would establish a public-private working group to address financial system cybersecurity risks posed by Mythos, Reuters reported.
Megabank Access
Japan’s three megabanks are expected to gain access to Mythos within approximately two weeks, according to a source cited by Reuters. This would make them among the first financial institutions outside the existing ~40 organizations in Anthropic’s Project Glasswing program to receive controlled access to the model’s vulnerability-discovery capabilities.
A New Governance Template
Japan’s approach is notable for its speed and structure. Within days of Mythos becoming a policy concern, the government convened cross-ministerial meetings, the ruling party delivered formal recommendations to the prime minister, and the finance minister raised it in bilateral talks with the U.S. Treasury. If Anthropic formalizes participation in the coalition, Japan would establish a model where the developer of a frontier AI agent’s vulnerability-discovery capabilities sits inside the government’s defensive planning process, contributing knowledge of the system’s characteristics and limitations directly to national policy.